stpm2_standsurv command; trouble with estimation of confidence intervals and standard errors

[Claire Lawson]

I would welcome advice on using the stpm2_standsurv command. When I include the permitted option 'ci' to estimate a confidence interval I get the error message below.

I tried to use the 'se' command instead to estimate the standard error and whilst I did not get an error message, no se was produced.

I used the following commands

. stpm2 age gender year_num CMno, scale(hazard) df(5) dftvc(4) eform

. stpm2_standsurv, atvars(male female) at1(gender 0) at2(gender 1) timevar(tt) ci

You can not use the ci option with the meansurv option
r(198);

Any suggestions much appreciated.

Thank you

[Paul Lambert]

Hi Claire,

I could not replicate this problem. Check you have the most up to date versions

. which stpm2
c:\ado\stbplus\s\stpm2.ado
*! version 1.7.2 02July2018

. which stpm2_standsurv
c:\ado\stbplus\s\stpm2_standsurv.ado
*! version 1.1.2 12Jun2018

Below is an example that works,

. webuse brcancer, clear
(German breast cancer data)

. stset rectime, f(censrec=1) scale(365.24)

     failure event:  censrec == 1
obs. time interval:  (0, rectime]
 exit on or before:  failure
    t for analysis:  time/365.24

------------------------------------------------------------------------------
        686  total observations
          0  exclusions
------------------------------------------------------------------------------
        686  observations remaining, representing
        299  failures in single-record/single-failure data
  2,112.036  total analysis time at risk and under observation
                                                at risk from t =         0
                                     earliest observed entry t =         0
                                          last observed exit t =  7.280145

. stpm2 hormon x1 x3, scale(hazard) df(4) tvc(hormon) dftvc(3)

Iteration 0:   log likelihood = -660.97597  
Iteration 1:   log likelihood = -660.63796  
Iteration 2:   log likelihood = -660.63627  
Iteration 3:   log likelihood = -660.63627  

Log likelihood = -660.63627                     Number of obs     =        686

--------------------------------------------------------------------------------
               |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
---------------+----------------------------------------------------------------
xb             |
        hormon |  -.4378755    .165474    -2.65   0.008    -.7621986   -.1135524
            x1 |   .0010953   .0061046     0.18   0.858    -.0108696    .0130601
            x3 |    .015271   .0035736     4.27   0.000     .0082669    .0222751
         _rcs1 |    1.48559   .1511487     9.83   0.000     1.189344    1.781836
         _rcs2 |   .4595725    .146068     3.15   0.002     .1732845    .7458606
         _rcs3 |   .0093231   .0511747     0.18   0.855    -.0909774    .1096236
         _rcs4 |  -.0372922   .0184427    -2.02   0.043    -.0734392   -.0011452
  _rcs_hormon1 |   .1536967   .3026641     0.51   0.612     -.439514    .7469075
  _rcs_hormon2 |   .0966158   .2695097     0.36   0.720    -.4316135    .6248451
  _rcs_hormon3 |  -.0069616   .0766581    -0.09   0.928    -.1572088    .1432855
         _cons |  -1.671857   .3523227    -4.75   0.000    -2.362396   -.9813169
--------------------------------------------------------------------------------

. range tt 0 5 100
(586 missing values generated)

. stpm2_standsurv, atvars(notrt trt) at1(hormon 0) at2(hormon 1) timevar(tt) ci

.

[Claire Lawson]

Thank you Paul, you were correct! I had a 2012 version of stpm2 installed. Now works perfectly, thank you.